Mailing List Archive

primary waits for secondary: why?
I cannot understand the following situation:

1) primary node_a goes down
2) secondary node_b takes over and now is primary
3) former primary node_a harddisk crashes during SyncAll after reboot
4) new primary node_b goes down due to power fail
5) power is back, node_b boots
6) node_a is not ok, harddisk crashed
7) node_b asks, if it should become primary
8) manually answered: yes
9) it does nothing still node_a boots also

I tested this twice with the same result, so I run in the following problem:
node_b was primary before power fail and so has consistent data, but I cannot use these data
until node_a has a new harddisk.
Imo this is not necessary, node_b should boot, even if node_a is not on, since it may take a day
to buy and install a new harddisk for node_a. node_b was primary and should and can become
primary after reboot, but it does not.
In the described case, I have not high availability but no availability. Or is there some reason for
this behavior which I cannot see? Any hint appreciated!




mfg ar

--
mailto:andreas@example.com
http://www.rittershofer.de
PGP-Public-Key http://www.rittershofer.de/ari.htm
RE: primary waits for secondary: why? [ In reply to ]
On 15 Nov 01, at 8:49, Michael Appeldorn wrote:

> Maybe your drbd.conf on node_b lacks a disk-size entry. my drbd-dev was
> lately unmountable without.

No, on both nodes there are the correct disk-size entries.

>
> And for point 9 : it does nothing if you misspell "yes"/"Yes" :O)_

Yes, I know. It was a correct "yes".


mfg ar

--
mailto:andreas@example.com
http://www.rittershofer.de
PGP-Public-Key http://www.rittershofer.de/ari.htm
RE: primary waits for secondary: why? [ In reply to ]
>I cannot understand the following situation:

>1) primary node_a goes down
>2) secondary node_b takes over and now is primary
>3) former primary node_a harddisk crashes during SyncAll after reboot
>4) new primary node_b goes down due to power fail
>5) power is back, node_b boots
>6) node_a is not ok, harddisk crashed
>7) node_b asks, if it should become primary
>8) manually answered: yes
>9) it does nothing still node_a boots also

>I tested this twice with the same result, so I run in the following
problem:
>node_b was primary before power fail and so has consistent data, but I
cannot use these data
>until node_a has a new harddisk.
>Imo this is not necessary, node_b should boot, even if node_a is not on,
since it may take a day
>to buy and install a new harddisk for node_a. node_b was primary and should
and can become
>primary after reboot, but it does not.
>In the described case, I have not high availability but no availability. Or
is there some reason for
>this behavior which I cannot see? Any hint appreciated!

Maybe your drbd.conf on node_b lacks a disk-size entry. my drbd-dev was
lately unmountable without.

And for point 9 : it does nothing if you misspell "yes"/"Yes" :O)_

Michael